lag length in time series analysis

[julia_uno]

hello
hope u can help.
I am operating with daily data. How many lags should I consider for a unit root test?
Autocorrelation (based on ACF and PCF) is significant for Lag 1,2 and also lag 15..should I inclulde lag 15? in daily data??
And also for my AR model - should I include lag 1,2 and 15? or only lag 1,2?

What is the test after I have constr. my AR model? Q-test and ARCH, right? so if I cannot reject the Ho of Q-stat my model is well specified (even without including lag 15) however if I have ARCH effects I need to constr. an ARCH model is that the right procedure?

Thanks.